What is known about the Magnetic North Pole's location before 1800?

I was recently startled to find that the Earth's North Magnetic Pole is moving at upwards of 40 km per year, with an additional ~80 km daily elliptical drift about its mean position due to variations in solar wind. The Wikipedia article on the NMP has a wealth of information, including this image:

I'm particularly surprised at the existence of "modelled" data, showing a lot of detail, about the NMP's location for centuries before anyone went there. Neither that Wikipedia article nor this one on the Earth's magnetic field, which include the image, mention where the modelled data come from.

I get the impression from sources like this that the model is derived from historical measurements of magnetic declination. If that is the case I would be disinclined to trust the model much: the Earth's magnetic field has a complex enough structure that the magnetic north and south are not antipodal, with nontrivial spatial variations in magnetic declination at any given time.

Has anyone seen references that explain this model and what measurements they are based on?

-
There are geological data showing variations in the earth's field going back over 5 million years. See en.wikipedia.org/wiki/Geomagnetic_reversal . I don't know how much this method tells you about the location of the pole, or whether it overlaps in time at all with historical records. It's not obvious to me that there has to be a single, well-defined point on the earth's surface at any given time at which the field is vertical. There could easily be more than one. – Ben Crowell May 12 '13 at 20:43

Here is the home page for the GUFM model website. It also includes a link to a freely available pdf of the modern reference. Also of interest is the NOAA WEBSITE.

GUFM MODEL HOMEPAGE

NOAA PAGE

Note that this model is based upon catalogs of geomagnetic field measurements, these did exist prior to 1800 - although as you would expect their quality and quantity decreases the farther back in time you go. The model is trying to solve an Inverse Problem. A 'solution' represents the model that provides the best fit to the available observations. It therefore may be considered our best empirical knowledge of the historic geomagnetic field based upon our limited observations.

Results such as the figure that you have posted would be more useful if, in addition to the most-likely values, they also displayed an objective measure of uncertainty (such as confidence limits) in the figure.

To clarify these terms a bit - This study is based on historic records (mostly ship's logs). Empirical information about the pre-historic geomagnetic field over geological time-scales is available to the science of paleomagnetism through the application of our understanding of rock magnetism. Rock magnetism is the study of the magnetic properties of rocks, sediments and soils. The field of rock magnetism arose out of the need in paleomagnetism to understand how rocks record the Earth's magnetic field.

-

Apologies for the hasty questioning. I found the reference in the File Description of the wikipedia image description page. The model is apparently called the GUFM model, for which a good reference is

Four centuries of geomagnetic secular variation from historical records. A Jackson, A R T Jonkers and M R Walker. Phil. Trans. R. Soc. Lond. A 358 no. 1768 (2000), pp.957-990, JSTOR 2666741.

That paper uses a spherical-harmonics spatial decomposition with a cutoff at $\ell=14$, which is enough to capture variations on a scale of some hundreds of kilometres, and more than enough to capture non-antipodal north and south poles.

I am happy to accept answers that delve into the details of that or related papers, and if I have time I will later expand this review.

-